Kingston HyperX DDR2-1067 1GB Dual Channel Kit

September 16, 2006
By admin

Product:
Kingston HyperX PC8500
Author: Paul Mercer
Category:
Enthusiast/Overclocker
Date: September 2006
Product cost: MSRP $284
($248.50 at allstarshop.com)
Manufacturer: Kingston
Spelling and Grammatical Editor:
Jason Hambly
This email address is being protected from
spam bots, you need Javascript enabled to view it


 

Introduction:

It?s been a while since we have had a memory review at PlanetX64;
this time round we have a kit from reputable manufacturer – Kingston.
We’ve
had a few Kingston products for review in the past and they have always been good
quality kits. This is Kingston?s fastest memory to date, their PC2-8500 HyperX DDR2.

Let?s see how well they perform and overclock…

{mospagebreak title= Packaging}

The Packaging of the Kingston PC8500 kit is both lightweight and
practical, easy to open to get the goodies out and it should be strong enough
to make it to your home in one piece (provided certain shipping companies don’t get their hands on it).

      


Included in the box is a small leaflet explaining the
correct precautions to take when handling and installing the modules.

  


 
{mospagebreak title= Design and Features}


The design of the module
matches what we have become used to seeing from Kingson?s HyperX line. Using
blue coloured heat spreaders emblazoned with the Kinston logo?s, these are
firmly attached using metal clips on the outside and thermal tape underneath. I decided
not to try removing the spreaders for fear of removing one of the memory chips or components with them.
The modules have been tested and certified to run at DDR2-1066 (533MHz) with
latency timings of 5-5-5-15 using just 2.2V. The SPD has been programmed to
JEDEC standard latency 800 MHz timing of 5-5-5-15 at 1.8V; this is to ensure
stability and compatibility with a wider range of motherboards.
 

FEATURES:

Power supply: Vdd: 1.8V ? 0.1V, Vddq: 1.8V ? 0.1V
Double-data-rate architecture; two data transfers per clock cycle
Bidirectional data strobe (DQS)
Differential clock inputs (CK and CK)
DLL aligns DQ and DQS transition with CK transition
Programmable Read latency 5 (clock)
Burst Length: 4, 8 (Interleave/nibble sequential)
Programmable Burst type (sequential & interleave)
Timing Reference: 5-5-5-15 at +1.8V / 5-5-5-15 at +2.2V
Edge aligned data output, centre aligned data input
Auto & Self refresh, 7.8us refresh interval (8K/64ms refresh)
Serial presence detect with EEPROM
High Performance Heat Spreader
PCB: Height 1.180? (30.00mm), single sided component

PERFORMANCE:

Clock Cycle Time (tCK) CL=5 2.5ns (min.) / 8ns (max.)
Row Cycle Time (tRC) 51.5ns (min.)
Refresh to Active/Refresh Command Time (tRFC) 105ns
Row Active Time (tRAS) 39ns (min.) / 70,000ns (max.)
Single Power Supply of +1.8V (+/- .1V)
Power 2.025 W (operating)
UL Rating 94 V – 0
Operating Temperature 0o C to 55o C
Storage Temperature -55o C to +125o C



 

{mospagebreak title=Performance}

Test System:

Intel Core 2 Duo E6700
Asus P5B Deluxe WiFi/AP BIOS:0706
1GB Kingston HyperX PC-8500 (2x512MB)
XFX 7900GT
1x Hitachi Deskstar 80GB 7200RPM SATA2 Hard Drive
Generic DVD Rom Drive
Tagan 580W EasyCon PSU
Windows XP Professional and Windows XP x64

As always, Windows was fully updated and a number of software
applications were installed onto the system and drives defragged before any
testing started.

Software installed included the following:
Avast Antivirus
Microsoft Office 2003
Diskeeper 10.0
Adobe Acrobat Reader 7.0
CyberLink Power DVD
Firefox
 

The Asus P5B Deluxe board with its newer 965P chipset allowed me to run
the memory at the specified speed without having to overclock the CPU.

Incredibly, this kit ran at the specified speed of DDR2-1067 with only
1.9V

Lavalys Everest (Win x86)

Lavalys Everest (Win x64)

 

SiSoftware Sandra 2007 (Win x86)

SiSoftware Sandra 2007 (Win x64)

Memory Bandwidth

Memory Bandwidth
 

Memory Latency

Memory Latency

 

Memtach (Win x86) 

Memtach (Win x64) 

 

 

 

RightMark Memory Analyser (Win x86)

RightMark Memory Analyser (Win x64)

Checksum

Checksum
 

Memory Bandwidth

Memory Bandwidth

Memory Walk

Memory Walk
 

Stream

Stream

Substring Search

Substring Search
 

 


Super Pi Mod (Win x86)
Super Pi Mod (Win x64)

Core 0
0h 21m 25.672s

Core 1
0h 21m 22.438s

Core 0
0h 20m 36.594s

Core 1
0h 20m 29.391


{mospagebreak title=Overclocking}

With Micron D9GKX chips on the PCB these kits are
already overclocked. Micron designed and specified the default speed of D9GKX
chips at DDR2-800, CAS 5, 1.8V. With this in mind, I was not expecting a huge
amount of overclocking room.

For the final overclocking tests and results, I used
a maximum VDimm of 2.30V. 2.35V would allow me to benchmark at DDR2-1260, but Memtest
would display errors after only a couple of hours of testing… this particular kit also
seemed to have a ?sweet-spot? at around 2.3V. I could also benchmark at 4-4-4-8 1067MHz, 2.35V, but since it was not memtest stable, those results will not be shown.

For the first test, I decided to try to
see how much I could tighten the latencies at the default speed of DDR2-1067.
Eventually I was able to get 4-5-4-8 stable, these tighter timings should reduce the
latency over the default timings.

DDR2-1067 4-5-4-8 & 2.3

Lavalys Everest (Win x86)

Lavalys Everest (Win x64)

 

SiSoftware Sandra 2007 (Win x86)

SiSoftware Sandra 2007 (Win x64)

Memory Bandwidth

Memory Bandwidth
 

Memory Latency
 

Memory Latency

 

Memtach (Win x86) 

Memtach (Win x64) 

 

 

 

RightMark Memory Analyser (Win x86)

RightMark Memory Analyser (Win x64)

Checksum

Checksum
 

Memory Bandwidth

Memory Bandwidth

Memory Walk

Memory Walk
 

Stream

Stream

Substring Search

Substring Search
 


Super Pi Mod (Win x86)
Super Pi Mod (Win x64)

Core 0

0h 21m 01.715s

Core 1

0h 20m 59.375

Core 0

0h 20m 13.000s

Core 1

0h 20m 06.656


When testing the speed limit of the default timings
( 5-5-5-15) and  raising the memory above
1067MHz it became necessary to overclock the CPU as well. Although, this was
unavoidable, I did reduce the CPU multiplier to x9 in an attempt to minimise
the affect of increased CPU speed.

I started increasing the FSB a couple of MHz at a
time and testing at each point. I was very surprised when the modules passed
550MHz and amazed when they passed 600MHz (DDR2-1200). Eventually though, they did
finally reach their limit with 5-5-5-15 timings and 2.30V at a final speed of 624MHz
or DDR2-1248, which was very impressive indeed.
We can see from the following results that the Core 2 Duo chips really enjoy the extra bandwidth made available by increasing the speed of the memory and FSB.

DDR2-1248 5-5-5-15 & 2.3V

Lavalys Everest (Win x86)

Lavalys Everest (Win x64)

 

SiSoftware Sandra 2007 (Win x86)

SiSoftware Sandra 2007 (Win x64)

Memory Bandwidth

Memory Bandwidth
 

Memory Latency
 

Memory Latency

 

Memtach (Win x86) 

Memtach (Win x64) 

 

 

 

RightMark Memory Analyser (Win x86)

RightMark Memory Analyser (Win x64)

Checksum

Checksum
 

Memory Bandwidth

Memory Bandwidth

Memory Walk

Memory Walk
 

Stream

Stream

Substring Search

Substring Search
 


Super Pi Mod (Win x86)
Super Pi Mod (Win x64)

Core 0

0h 20m 12.344s

Core 1

0h 20m 10.672s

Core 0

0h 18m 57.766s

Core 1

0h 18m 52.188s



All results shown in this review are stable results.
I occasionally see memory reviews around the web that display all manner of
amazing results and timings, but these are often only stable enough to run the
benchmark shown. I do not really like testing like this and to me, stability
means more than fast benchmarks (although stability and speed together is
nice). So as well as having to pass each benchmark, final speeds were tested
using HCI Memtest for a minimum of 6 hours to ensure the results I obtained to be
not just benchmark stable.

  

{mospagebreak title= Price and Warranty}

 
Looking at prices of other memory kits with similar specifications, it
seems that the Kingston HyperX can be found for roughly the same price. Higher
prices on PC8500 memory can be attributed to the more stringent speed binning
process that takes place.

Kingston Memory modules including ValueRAM, HyperX and Kingston system
specific memory, Flash memory cards (e.g. Secure Digital, CompactFlash,
MultiMediaCard, SmartMedia) ATA Flash, and Linear Flash; memory
expansion-boards, networking adapters; routers, hubs and switches without cooling
fans (excluding the power supply), Flash adapters and microprocessor upgrade
products are all covered by a Limited Lifetime Warranty. For more details, please
visit

http://www.kingston.com/company/warranty.asp


 

{mospagebreak title=Conclusion}

All I can say about these modules is ?WOW?
I expected a little
overclocking room, but nowhere near as much as I got. This was a nice
introduction for me into overclocking DDR2 and I would highly recommend these
modules to anyone looking for a really overclockable 1GB kit….
Unfortunately these days not a lot of people are interested in buying 1GB kits anymore and want 2GB or more
in their PC’s. Kingston also produces a 2GB kit of HyperX PC-8500 memory,
but I was unable to find prices at this time.
Although the saying goes ?you get what you pay for?, to some the price
may seem a little high, especially with the recent DDR2 price increases. As mentioned previously though, a lot more work goes into
producing high speed memory such as this and of course you get a lifetime guarantee and support from Kingston.
These Kingston HyperX PC8500 kits would certainly suit the overclockers
out there. Capable of DDR2-1200 and above, these modules were very flexible and easy to work with, if it weren’t for the high price, these sticks would have undoubtedly won a Best on the Planet award.

 

Pros

  • High Overclocks
  • Kingston brand and support
  • Flexible Latencies

Cons

  • 1Gb kit
  • High Price

 

Specifications/Features: 4.5/5
Appearance: 4.25/5
Performance: 4.5/5
Price/Warranty: 4/5

Overall: 17.25/20

 

 We would like to thank Claudia Martinez at Kingston for our review sample.

Leave a Reply

Your email address will not be published. Required fields are marked *

*